summaryrefslogtreecommitdiffstats
path: root/arch/powerpc/platforms/powernv/npu-dma.c
blob: 8cdf91f5d3a4ec30d279924e1722d480b4c655d9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
/*
 * This file implements the DMA operations for NVLink devices. The NPU
 * devices all point to the same iommu table as the parent PCI device.
 *
 * Copyright Alistair Popple, IBM Corporation 2015.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 */

#include <linux/slab.h>
#include <linux/mmu_notifier.h>
#include <linux/mmu_context.h>
#include <linux/of.h>
#include <linux/export.h>
#include <linux/pci.h>
#include <linux/memblock.h>
#include <linux/iommu.h>

#include <asm/tlb.h>
#include <asm/powernv.h>
#include <asm/reg.h>
#include <asm/opal.h>
#include <asm/io.h>
#include <asm/iommu.h>
#include <asm/pnv-pci.h>
#include <asm/msi_bitmap.h>
#include <asm/opal.h>

#include "powernv.h"
#include "pci.h"

#define npu_to_phb(x) container_of(x, struct pnv_phb, npu)

/*
 * spinlock to protect initialisation of an npu_context for a particular
 * mm_struct.
 */
static DEFINE_SPINLOCK(npu_context_lock);

/*
 * When an address shootdown range exceeds this threshold we invalidate the
 * entire TLB on the GPU for the given PID rather than each specific address in
 * the range.
 */
#define ATSD_THRESHOLD (2*1024*1024)

/*
 * Other types of TCE cache invalidation are not functional in the
 * hardware.
 */
static struct pci_dev *get_pci_dev(struct device_node *dn)
{
	struct pci_dn *pdn = PCI_DN(dn);

	return pci_get_domain_bus_and_slot(pci_domain_nr(pdn->phb->bus),
					   pdn->busno, pdn->devfn);
}

/* Given a NPU device get the associated PCI device. */
struct pci_dev *pnv_pci_get_gpu_dev(struct pci_dev *npdev)
{
	struct device_node *dn;
	struct pci_dev *gpdev;

	if (WARN_ON(!npdev))
		return NULL;

	if (WARN_ON(!npdev->dev.of_node))
		return NULL;

	/* Get assoicated PCI device */
	dn = of_parse_phandle(npdev->dev.of_node, "ibm,gpu", 0);
	if (!dn)
		return NULL;

	gpdev = get_pci_dev(dn);
	of_node_put(dn);

	return gpdev;
}
EXPORT_SYMBOL(pnv_pci_get_gpu_dev);

/* Given the real PCI device get a linked NPU device. */
struct pci_dev *pnv_pci_get_npu_dev(struct pci_dev *gpdev, int index)
{
	struct device_node *dn;
	struct pci_dev *npdev;

	if (WARN_ON(!gpdev))
		return NULL;

	/* Not all PCI devices have device-tree nodes */
	if (!gpdev->dev.of_node)
		return NULL;

	/* Get assoicated PCI device */
	dn = of_parse_phandle(gpdev->dev.of_node, "ibm,npu", index);
	if (!dn)
		return NULL;

	npdev = get_pci_dev(dn);
	of_node_put(dn);

	return npdev;
}
EXPORT_SYMBOL(pnv_pci_get_npu_dev);

#define NPU_DMA_OP_UNSUPPORTED()					\
	dev_err_once(dev, "%s operation unsupported for NVLink devices\n", \
		__func__)

static void *dma_npu_alloc(struct device *dev, size_t size,
			   dma_addr_t *dma_handle, gfp_t flag,
			   unsigned long attrs)
{
	NPU_DMA_OP_UNSUPPORTED();
	return NULL;
}

static void dma_npu_free(struct device *dev, size_t size,
			 void *vaddr, dma_addr_t dma_handle,
			 unsigned long attrs)
{
	NPU_DMA_OP_UNSUPPORTED();
}

static dma_addr_t dma_npu_map_page(struct device *dev, struct page *page,
				   unsigned long offset, size_t size,
				   enum dma_data_direction direction,
				   unsigned long attrs)
{
	NPU_DMA_OP_UNSUPPORTED();
	return 0;
}

static int dma_npu_map_sg(struct device *dev, struct scatterlist *sglist,
			  int nelems, enum dma_data_direction direction,
			  unsigned long attrs)
{
	NPU_DMA_OP_UNSUPPORTED();
	return 0;
}

static int dma_npu_dma_supported(struct device *dev, u64 mask)
{
	NPU_DMA_OP_UNSUPPORTED();
	return 0;
}

static u64 dma_npu_get_required_mask(struct device *dev)
{
	NPU_DMA_OP_UNSUPPORTED();
	return 0;
}

static const struct dma_map_ops dma_npu_ops = {
	.map_page		= dma_npu_map_page,
	.map_sg			= dma_npu_map_sg,
	.alloc			= dma_npu_alloc,
	.free			= dma_npu_free,
	.dma_supported		= dma_npu_dma_supported,
	.get_required_mask	= dma_npu_get_required_mask,
};

/*
 * Returns the PE assoicated with the PCI device of the given
 * NPU. Returns the linked pci device if pci_dev != NULL.
 */
static struct pnv_ioda_pe *get_gpu_pci_dev_and_pe(struct pnv_ioda_pe *npe,
						  struct pci_dev **gpdev)
{
	struct pnv_phb *phb;
	struct pci_controller *hose;
	struct pci_dev *pdev;
	struct pnv_ioda_pe *pe;
	struct pci_dn *pdn;

	pdev = pnv_pci_get_gpu_dev(npe->pdev);
	if (!pdev)
		return NULL;

	pdn = pci_get_pdn(pdev);
	if (WARN_ON(!pdn || pdn->pe_number == IODA_INVALID_PE))
		return NULL;

	hose = pci_bus_to_host(pdev->bus);
	phb = hose->private_data;
	pe = &phb->ioda.pe_array[pdn->pe_number];

	if (gpdev)
		*gpdev = pdev;

	return pe;
}

long pnv_npu_set_window(struct pnv_ioda_pe *npe, int num,
		struct iommu_table *tbl)
{
	struct pnv_phb *phb = npe->phb;
	int64_t rc;
	const unsigned long size = tbl->it_indirect_levels ?
		tbl->it_level_size : tbl->it_size;
	const __u64 start_addr = tbl->it_offset << tbl->it_page_shift;
	const __u64 win_size = tbl->it_size << tbl->it_page_shift;

	pe_info(npe, "Setting up window %llx..%llx pg=%lx\n",
			start_addr, start_addr + win_size - 1,
			IOMMU_PAGE_SIZE(tbl));

	rc = opal_pci_map_pe_dma_window(phb->opal_id,
			npe->pe_number,
			npe->pe_number,
			tbl->it_indirect_levels + 1,
			__pa(tbl->it_base),
			size << 3,
			IOMMU_PAGE_SIZE(tbl));
	if (rc) {
		pe_err(npe, "Failed to configure TCE table, err %lld\n", rc);
		return rc;
	}
	pnv_pci_ioda2_tce_invalidate_entire(phb, false);

	/* Add the table to the list so its TCE cache will get invalidated */
	pnv_pci_link_table_and_group(phb->hose->node, num,
			tbl, &npe->table_group);

	return 0;
}

long pnv_npu_unset_window(struct pnv_ioda_pe *npe, int num)
{
	struct pnv_phb *phb = npe->phb;
	int64_t rc;

	pe_info(npe, "Removing DMA window\n");

	rc = opal_pci_map_pe_dma_window(phb->opal_id, npe->pe_number,
			npe->pe_number,
			0/* levels */, 0/* table address */,
			0/* table size */, 0/* page size */);
	if (rc) {
		pe_err(npe, "Unmapping failed, ret = %lld\n", rc);
		return rc;
	}
	pnv_pci_ioda2_tce_invalidate_entire(phb, false);

	pnv_pci_unlink_table_and_group(npe->table_group.tables[num],
			&npe->table_group);

	return 0;
}

/*
 * Enables 32 bit DMA on NPU.
 */
static void pnv_npu_dma_set_32(struct pnv_ioda_pe *npe)
{
	struct pci_dev *gpdev;
	struct pnv_ioda_pe *gpe;
	int64_t rc;

	/*
	 * Find the assoicated PCI devices and get the dma window
	 * information from there.
	 */
	if (!npe->pdev || !(npe->flags & PNV_IODA_PE_DEV))
		return;

	gpe = get_gpu_pci_dev_and_pe(npe, &gpdev);
	if (!gpe)
		return;

	rc = pnv_npu_set_window(npe, 0, gpe->table_group.tables[0]);

	/*
	 * We don't initialise npu_pe->tce32_table as we always use
	 * dma_npu_ops which are nops.
	 */
	set_dma_ops(&npe->pdev->dev, &dma_npu_ops);
}

/*
 * Enables bypass mode on the NPU. The NPU only supports one
 * window per link, so bypass needs to be explicitly enabled or
 * disabled. Unlike for a PHB3 bypass and non-bypass modes can't be
 * active at the same time.
 */
static int pnv_npu_dma_set_bypass(struct pnv_ioda_pe *npe)
{
	struct pnv_phb *phb = npe->phb;
	int64_t rc = 0;
	phys_addr_t top = memblock_end_of_DRAM();

	if (phb->type != PNV_PHB_NPU_NVLINK || !npe->pdev)
		return -EINVAL;

	rc = pnv_npu_unset_window(npe, 0);
	if (rc != OPAL_SUCCESS)
		return rc;

	/* Enable the bypass window */

	top = roundup_pow_of_two(top);
	dev_info(&npe->pdev->dev, "Enabling bypass for PE %x\n",
			npe->pe_number);
	rc = opal_pci_map_pe_dma_window_real(phb->opal_id,
			npe->pe_number, npe->pe_number,
			0 /* bypass base */, top);

	if (rc == OPAL_SUCCESS)
		pnv_pci_ioda2_tce_invalidate_entire(phb, false);

	return rc;
}

void pnv_npu_try_dma_set_bypass(struct pci_dev *gpdev, bool bypass)
{
	int i;
	struct pnv_phb *phb;
	struct pci_dn *pdn;
	struct pnv_ioda_pe *npe;
	struct pci_dev *npdev;

	for (i = 0; ; ++i) {
		npdev = pnv_pci_get_npu_dev(gpdev, i);

		if (!npdev)
			break;

		pdn = pci_get_pdn(npdev);
		if (WARN_ON(!pdn || pdn->pe_number == IODA_INVALID_PE))
			return;

		phb = pci_bus_to_host(npdev->bus)->private_data;

		/* We only do bypass if it's enabled on the linked device */
		npe = &phb->ioda.pe_array[pdn->pe_number];

		if (bypass) {
			dev_info(&npdev->dev,
					"Using 64-bit DMA iommu bypass\n");
			pnv_npu_dma_set_bypass(npe);
		} else {
			dev_info(&npdev->dev, "Using 32-bit DMA via iommu\n");
			pnv_npu_dma_set_32(npe);
		}
	}
}

/* Switch ownership from platform code to external user (e.g. VFIO) */
void pnv_npu_take_ownership(struct pnv_ioda_pe *npe)
{
	struct pnv_phb *phb = npe->phb;
	int64_t rc;

	/*
	 * Note: NPU has just a single TVE in the hardware which means that
	 * while used by the kernel, it can have either 32bit window or
	 * DMA bypass but never both. So we deconfigure 32bit window only
	 * if it was enabled at the moment of ownership change.
	 */
	if (npe->table_group.tables[0]) {
		pnv_npu_unset_window(npe, 0);
		return;
	}

	/* Disable bypass */
	rc = opal_pci_map_pe_dma_window_real(phb->opal_id,
			npe->pe_number, npe->pe_number,
			0 /* bypass base */, 0);
	if (rc) {
		pe_err(npe, "Failed to disable bypass, err %lld\n", rc);
		return;
	}
	pnv_pci_ioda2_tce_invalidate_entire(npe->phb, false);
}

struct pnv_ioda_pe *pnv_pci_npu_setup_iommu(struct pnv_ioda_pe *npe)
{
	struct pnv_phb *phb = npe->phb;
	struct pci_bus *pbus = phb->hose->bus;
	struct pci_dev *npdev, *gpdev = NULL, *gptmp;
	struct pnv_ioda_pe *gpe = get_gpu_pci_dev_and_pe(npe, &gpdev);

	if (!gpe || !gpdev)
		return NULL;

	list_for_each_entry(npdev, &pbus->devices, bus_list) {
		gptmp = pnv_pci_get_gpu_dev(npdev);

		if (gptmp != gpdev)
			continue;

		pe_info(gpe, "Attached NPU %s\n", dev_name(&npdev->dev));
		iommu_group_add_device(gpe->table_group.group, &npdev->dev);
	}

	return gpe;
}

/* Maximum number of nvlinks per npu */
#define NV_MAX_LINKS 6

/* Maximum index of npu2 hosts in the system. Always < NV_MAX_NPUS */
static int max_npu2_index;

struct npu_context {
	struct mm_struct *mm;
	struct pci_dev *npdev[NV_MAX_NPUS][NV_MAX_LINKS];
	struct mmu_notifier mn;
	struct kref kref;
	bool nmmu_flush;

	/* Callback to stop translation requests on a given GPU */
	void (*release_cb)(struct npu_context *context, void *priv);

	/*
	 * Private pointer passed to the above callback for usage by
	 * device drivers.
	 */
	void *priv;
};

struct mmio_atsd_reg {
	struct npu *npu;
	int reg;
};

/*
 * Find a free MMIO ATSD register and mark it in use. Return -ENOSPC
 * if none are available.
 */
static int get_mmio_atsd_reg(struct npu *npu)
{
	int i;

	for (i = 0; i < npu->mmio_atsd_count; i++) {
		if (!test_and_set_bit_lock(i, &npu->mmio_atsd_usage))
			return i;
	}

	return -ENOSPC;
}

static void put_mmio_atsd_reg(struct npu *npu, int reg)
{
	clear_bit_unlock(reg, &npu->mmio_atsd_usage);
}

/* MMIO ATSD register offsets */
#define XTS_ATSD_AVA  1
#define XTS_ATSD_STAT 2

static void mmio_launch_invalidate(struct mmio_atsd_reg *mmio_atsd_reg,
				unsigned long launch, unsigned long va)
{
	struct npu *npu = mmio_atsd_reg->npu;
	int reg = mmio_atsd_reg->reg;

	__raw_writeq_be(va, npu->mmio_atsd_regs[reg] + XTS_ATSD_AVA);
	eieio();
	__raw_writeq_be(launch, npu->mmio_atsd_regs[reg]);
}

static void mmio_invalidate_pid(struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS],
				unsigned long pid, bool flush)
{
	int i;
	unsigned long launch;

	for (i = 0; i <= max_npu2_index; i++) {
		if (mmio_atsd_reg[i].reg < 0)
			continue;

		/* IS set to invalidate matching PID */
		launch = PPC_BIT(12);

		/* PRS set to process-scoped */
		launch |= PPC_BIT(13);

		/* AP */
		launch |= (u64)
			mmu_get_ap(mmu_virtual_psize) << PPC_BITLSHIFT(17);

		/* PID */
		launch |= pid << PPC_BITLSHIFT(38);

		/* No flush */
		launch |= !flush << PPC_BITLSHIFT(39);

		/* Invalidating the entire process doesn't use a va */
		mmio_launch_invalidate(&mmio_atsd_reg[i], launch, 0);
	}
}

static void mmio_invalidate_va(struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS],
			unsigned long va, unsigned long pid, bool flush)
{
	int i;
	unsigned long launch;

	for (i = 0; i <= max_npu2_index; i++) {
		if (mmio_atsd_reg[i].reg < 0)
			continue;

		/* IS set to invalidate target VA */
		launch = 0;

		/* PRS set to process scoped */
		launch |= PPC_BIT(13);

		/* AP */
		launch |= (u64)
			mmu_get_ap(mmu_virtual_psize) << PPC_BITLSHIFT(17);

		/* PID */
		launch |= pid << PPC_BITLSHIFT(38);

		/* No flush */
		launch |= !flush << PPC_BITLSHIFT(39);

		mmio_launch_invalidate(&mmio_atsd_reg[i], launch, va);
	}
}

#define mn_to_npu_context(x) container_of(x, struct npu_context, mn)

static void mmio_invalidate_wait(
	struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
{
	struct npu *npu;
	int i, reg;

	/* Wait for all invalidations to complete */
	for (i = 0; i <= max_npu2_index; i++) {
		if (mmio_atsd_reg[i].reg < 0)
			continue;

		/* Wait for completion */
		npu = mmio_atsd_reg[i].npu;
		reg = mmio_atsd_reg[i].reg;
		while (__raw_readq(npu->mmio_atsd_regs[reg] + XTS_ATSD_STAT))
			cpu_relax();
	}
}

/*
 * Acquires all the address translation shootdown (ATSD) registers required to
 * launch an ATSD on all links this npu_context is active on.
 */
static void acquire_atsd_reg(struct npu_context *npu_context,
			struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
{
	int i, j;
	struct npu *npu;
	struct pci_dev *npdev;
	struct pnv_phb *nphb;

	for (i = 0; i <= max_npu2_index; i++) {
		mmio_atsd_reg[i].reg = -1;
		for (j = 0; j < NV_MAX_LINKS; j++) {
			/*
			 * There are no ordering requirements with respect to
			 * the setup of struct npu_context, but to ensure
			 * consistent behaviour we need to ensure npdev[][] is
			 * only read once.
			 */
			npdev = READ_ONCE(npu_context->npdev[i][j]);
			if (!npdev)
				continue;

			nphb = pci_bus_to_host(npdev->bus)->private_data;
			npu = &nphb->npu;
			mmio_atsd_reg[i].npu = npu;
			mmio_atsd_reg[i].reg = get_mmio_atsd_reg(npu);
			while (mmio_atsd_reg[i].reg < 0) {
				mmio_atsd_reg[i].reg = get_mmio_atsd_reg(npu);
				cpu_relax();
			}
			break;
		}
	}
}

/*
 * Release previously acquired ATSD registers. To avoid deadlocks the registers
 * must be released in the same order they were acquired above in
 * acquire_atsd_reg.
 */
static void release_atsd_reg(struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
{
	int i;

	for (i = 0; i <= max_npu2_index; i++) {
		/*
		 * We can't rely on npu_context->npdev[][] being the same here
		 * as when acquire_atsd_reg() was called, hence we use the
		 * values stored in mmio_atsd_reg during the acquire phase
		 * rather than re-reading npdev[][].
		 */
		if (mmio_atsd_reg[i].reg < 0)
			continue;

		put_mmio_atsd_reg(mmio_atsd_reg[i].npu, mmio_atsd_reg[i].reg);
	}
}

/*
 * Invalidate either a single address or an entire PID depending on
 * the value of va.
 */
static void mmio_invalidate(struct npu_context *npu_context, int va,
			unsigned long address, bool flush)
{
	struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS];
	unsigned long pid = npu_context->mm->context.id;

	if (npu_context->nmmu_flush)
		/*
		 * Unfortunately the nest mmu does not support flushing specific
		 * addresses so we have to flush the whole mm once before
		 * shooting down the GPU translation.
		 */
		flush_all_mm(npu_context->mm);

	/*
	 * Loop over all the NPUs this process is active on and launch
	 * an invalidate.
	 */
	acquire_atsd_reg(npu_context, mmio_atsd_reg);
	if (va)
		mmio_invalidate_va(mmio_atsd_reg, address, pid, flush);
	else
		mmio_invalidate_pid(mmio_atsd_reg, pid, flush);

	mmio_invalidate_wait(mmio_atsd_reg);
	if (flush) {
		/*
		 * The GPU requires two flush ATSDs to ensure all entries have
		 * been flushed. We use PID 0 as it will never be used for a
		 * process on the GPU.
		 */
		mmio_invalidate_pid(mmio_atsd_reg, 0, true);
		mmio_invalidate_wait(mmio_atsd_reg);
		mmio_invalidate_pid(mmio_atsd_reg, 0, true);
		mmio_invalidate_wait(mmio_atsd_reg);
	}
	release_atsd_reg(mmio_atsd_reg);
}

static void pnv_npu2_mn_release(struct mmu_notifier *mn,
				struct mm_struct *mm)
{
	struct npu_context *npu_context = mn_to_npu_context(mn);

	/* Call into device driver to stop requests to the NMMU */
	if (npu_context->release_cb)
		npu_context->release_cb(npu_context, npu_context->priv);

	/*
	 * There should be no more translation requests for this PID, but we
	 * need to ensure any entries for it are removed from the TLB.
	 */
	mmio_invalidate(npu_context, 0, 0, true);
}

static void pnv_npu2_mn_change_pte(struct mmu_notifier *mn,
				struct mm_struct *mm,
				unsigned long address,
				pte_t pte)
{
	struct npu_context *npu_context = mn_to_npu_context(mn);

	mmio_invalidate(npu_context, 1, address, true);
}

static void pnv_npu2_mn_invalidate_range(struct mmu_notifier *mn,
					struct mm_struct *mm,
					unsigned long start, unsigned long end)
{
	struct npu_context *npu_context = mn_to_npu_context(mn);
	unsigned long address;

	if (end - start > ATSD_THRESHOLD) {
		/*
		 * Just invalidate the entire PID if the address range is too
		 * large.
		 */
		mmio_invalidate(npu_context, 0, 0, true);
	} else {
		for (address = start; address < end; address += PAGE_SIZE)
			mmio_invalidate(npu_context, 1, address, false);

		/* Do the flush only on the final addess == end */
		mmio_invalidate(npu_context, 1, address, true);
	}
}

static const struct mmu_notifier_ops nv_nmmu_notifier_ops = {
	.release = pnv_npu2_mn_release,
	.change_pte = pnv_npu2_mn_change_pte,
	.invalidate_range = pnv_npu2_mn_invalidate_range,
};

/*
 * Call into OPAL to setup the nmmu context for the current task in
 * the NPU. This must be called to setup the context tables before the
 * GPU issues ATRs. pdev should be a pointed to PCIe GPU device.
 *
 * A release callback should be registered to allow a device driver to
 * be notified that it should not launch any new translation requests
 * as the final TLB invalidate is about to occur.
 *
 * Returns an error if there no contexts are currently available or a
 * npu_context which should be passed to pnv_npu2_handle_fault().
 *
 * mmap_sem must be held in write mode and must not be called from interrupt
 * context.
 */
struct npu_context *pnv_npu2_init_context(struct pci_dev *gpdev,
			unsigned long flags,
			void (*cb)(struct npu_context *, void *),
			void *priv)
{
	int rc;
	u32 nvlink_index;
	struct device_node *nvlink_dn;
	struct mm_struct *mm = current->mm;
	struct pnv_phb *nphb;
	struct npu *npu;
	struct npu_context *npu_context;

	/*
	 * At present we don't support GPUs connected to multiple NPUs and I'm
	 * not sure the hardware does either.
	 */
	struct pci_dev *npdev = pnv_pci_get_npu_dev(gpdev, 0);

	if (!firmware_has_feature(FW_FEATURE_OPAL))
		return ERR_PTR(-ENODEV);

	if (!npdev)
		/* No nvlink associated with this GPU device */
		return ERR_PTR(-ENODEV);

	nvlink_dn = of_parse_phandle(npdev->dev.of_node, "ibm,nvlink", 0);
	if (WARN_ON(of_property_read_u32(nvlink_dn, "ibm,npu-link-index",
							&nvlink_index)))
		return ERR_PTR(-ENODEV);

	if (!mm || mm->context.id == 0) {
		/*
		 * Kernel thread contexts are not supported and context id 0 is
		 * reserved on the GPU.
		 */
		return ERR_PTR(-EINVAL);
	}

	nphb = pci_bus_to_host(npdev->bus)->private_data;
	npu = &nphb->npu;

	/*
	 * Setup the NPU context table for a particular GPU. These need to be
	 * per-GPU as we need the tables to filter ATSDs when there are no
	 * active contexts on a particular GPU. It is safe for these to be
	 * called concurrently with destroy as the OPAL call takes appropriate
	 * locks and refcounts on init/destroy.
	 */
	rc = opal_npu_init_context(nphb->opal_id, mm->context.id, flags,
				PCI_DEVID(gpdev->bus->number, gpdev->devfn));
	if (rc < 0)
		return ERR_PTR(-ENOSPC);

	/*
	 * We store the npu pci device so we can more easily get at the
	 * associated npus.
	 */
	spin_lock(&npu_context_lock);
	npu_context = mm->context.npu_context;
	if (npu_context) {
		if (npu_context->release_cb != cb ||
			npu_context->priv != priv) {
			spin_unlock(&npu_context_lock);
			opal_npu_destroy_context(nphb->opal_id, mm->context.id,
						PCI_DEVID(gpdev->bus->number,
							gpdev->devfn));
			return ERR_PTR(-EINVAL);
		}

		WARN_ON(!kref_get_unless_zero(&npu_context->kref));
	}
	spin_unlock(&npu_context_lock);

	if (!npu_context) {
		/*
		 * We can set up these fields without holding the
		 * npu_context_lock as the npu_context hasn't been returned to
		 * the caller meaning it can't be destroyed. Parallel allocation
		 * is protected against by mmap_sem.
		 */
		rc = -ENOMEM;
		npu_context = kzalloc(sizeof(struct npu_context), GFP_KERNEL);
		if (npu_context) {
			kref_init(&npu_context->kref);
			npu_context->mm = mm;
			npu_context->mn.ops = &nv_nmmu_notifier_ops;
			rc = __mmu_notifier_register(&npu_context->mn, mm);
		}

		if (rc) {
			kfree(npu_context);
			opal_npu_destroy_context(nphb->opal_id, mm->context.id,
					PCI_DEVID(gpdev->bus->number,
						gpdev->devfn));
			return ERR_PTR(rc);
		}

		mm->context.npu_context = npu_context;
	}

	npu_context->release_cb = cb;
	npu_context->priv = priv;

	/*
	 * npdev is a pci_dev pointer setup by the PCI code. We assign it to
	 * npdev[][] to indicate to the mmu notifiers that an invalidation
	 * should also be sent over this nvlink. The notifiers don't use any
	 * other fields in npu_context, so we just need to ensure that when they
	 * deference npu_context->npdev[][] it is either a valid pointer or
	 * NULL.
	 */
	WRITE_ONCE(npu_context->npdev[npu->index][nvlink_index], npdev);

	if (!nphb->npu.nmmu_flush) {
		/*
		 * If we're not explicitly flushing ourselves we need to mark
		 * the thread for global flushes
		 */
		npu_context->nmmu_flush = false;
		mm_context_add_copro(mm);
	} else
		npu_context->nmmu_flush = true;

	return npu_context;
}
EXPORT_SYMBOL(pnv_npu2_init_context);

static void pnv_npu2_release_context(struct kref *kref)
{
	struct npu_context *npu_context =
		container_of(kref, struct npu_context, kref);

	if (!npu_context->nmmu_flush)
		mm_context_remove_copro(npu_context->mm);

	npu_context->mm->context.npu_context = NULL;
}

/*
 * Destroy a context on the given GPU. May free the npu_context if it is no
 * longer active on any GPUs. Must not be called from interrupt context.
 */
void pnv_npu2_destroy_context(struct npu_context *npu_context,
			struct pci_dev *gpdev)
{
	int removed;
	struct pnv_phb *nphb;
	struct npu *npu;
	struct pci_dev *npdev = pnv_pci_get_npu_dev(gpdev, 0);
	struct device_node *nvlink_dn;
	u32 nvlink_index;

	if (WARN_ON(!npdev))
		return;

	if (!firmware_has_feature(FW_FEATURE_OPAL))
		return;

	nphb = pci_bus_to_host(npdev->bus)->private_data;
	npu = &nphb->npu;
	nvlink_dn = of_parse_phandle(npdev->dev.of_node, "ibm,nvlink", 0);
	if (WARN_ON(of_property_read_u32(nvlink_dn, "ibm,npu-link-index",
							&nvlink_index)))
		return;
	WRITE_ONCE(npu_context->npdev[npu->index][nvlink_index], NULL);
	opal_npu_destroy_context(nphb->opal_id, npu_context->mm->context.id,
				PCI_DEVID(gpdev->bus->number, gpdev->devfn));
	spin_lock(&npu_context_lock);
	removed = kref_put(&npu_context->kref, pnv_npu2_release_context);
	spin_unlock(&npu_context_lock);

	/*
	 * We need to do this outside of pnv_npu2_release_context so that it is
	 * outside the spinlock as mmu_notifier_destroy uses SRCU.
	 */
	if (removed) {
		mmu_notifier_unregister(&npu_context->mn,
					npu_context->mm);

		kfree(npu_context);
	}

}
EXPORT_SYMBOL(pnv_npu2_destroy_context);

/*
 * Assumes mmap_sem is held for the contexts associated mm.
 */
int pnv_npu2_handle_fault(struct npu_context *context, uintptr_t *ea,
			unsigned long *flags, unsigned long *status, int count)
{
	u64 rc = 0, result = 0;
	int i, is_write;
	struct page *page[1];

	/* mmap_sem should be held so the struct_mm must be present */
	struct mm_struct *mm = context->mm;

	if (!firmware_has_feature(FW_FEATURE_OPAL))
		return -ENODEV;

	WARN_ON(!rwsem_is_locked(&mm->mmap_sem));

	for (i = 0; i < count; i++) {
		is_write = flags[i] & NPU2_WRITE;
		rc = get_user_pages_remote(NULL, mm, ea[i], 1,
					is_write ? FOLL_WRITE : 0,
					page, NULL, NULL);

		/*
		 * To support virtualised environments we will have to do an
		 * access to the page to ensure it gets faulted into the
		 * hypervisor. For the moment virtualisation is not supported in
		 * other areas so leave the access out.
		 */
		if (rc != 1) {
			status[i] = rc;
			result = -EFAULT;
			continue;
		}

		status[i] = 0;
		put_page(page[0]);
	}

	return result;
}
EXPORT_SYMBOL(pnv_npu2_handle_fault);

int pnv_npu2_init(struct pnv_phb *phb)
{
	unsigned int i;
	u64 mmio_atsd;
	struct device_node *dn;
	struct pci_dev *gpdev;
	static int npu_index;
	uint64_t rc = 0;

	phb->npu.nmmu_flush =
		of_property_read_bool(phb->hose->dn, "ibm,nmmu-flush");
	for_each_child_of_node(phb->hose->dn, dn) {
		gpdev = pnv_pci_get_gpu_dev(get_pci_dev(dn));
		if (gpdev) {
			rc = opal_npu_map_lpar(phb->opal_id,
				PCI_DEVID(gpdev->bus->number, gpdev->devfn),
				0, 0);
			if (rc)
				dev_err(&gpdev->dev,
					"Error %lld mapping device to LPAR\n",
					rc);
		}
	}

	for (i = 0; !of_property_read_u64_index(phb->hose->dn, "ibm,mmio-atsd",
							i, &mmio_atsd); i++)
		phb->npu.mmio_atsd_regs[i] = ioremap(mmio_atsd, 32);

	pr_info("NPU%lld: Found %d MMIO ATSD registers", phb->opal_id, i);
	phb->npu.mmio_atsd_count = i;
	phb->npu.mmio_atsd_usage = 0;
	npu_index++;
	if (WARN_ON(npu_index >= NV_MAX_NPUS))
		return -ENOSPC;
	max_npu2_index = npu_index;
	phb->npu.index = npu_index;

	return 0;
}